Adaptive Antenna Array Processing with Kernels

Adaptive array processing improves performance of receivers in time varying scenarios. Linear methods provide easy to implement, adaptive and low computational burden solutions. For that reason, methods as LMS and RES are widely used in practice. Nonlinear algorithms are usually impractical due to their high complexity and because they often need large training datasets for their parameters to be adjusted. Also, it may be hard to obtain adaptive versions of them. Kernel methods, together with Support Vector Machines provide nonlinear, low complex machines that have good generalization performance. Here we present an adaptive nonlinear and complex-valued SVM that is applied to antenna array processing, and we compare it with the standard methods in several multiuser varying scenarios.